Vol. 584: 45–65, 2017 MARINE ECOLOGY PROGRESS SERIES Published December 7 https://doi.org/10.3354/meps12347 Mar Ecol Prog Ser

OPENPEN ACCESSCCESS

Distributions of krill and Antarctic silverfish and correlations with environmental variables in the western Ross Sea, Antarctica

L. Brynn Davis1,*, Eileen E. Hofmann1, John M. Klinck1, Andrea Piñones2,3,4, Michael S. Dinniman1

1Center for Coastal Physical Oceanography, Old Dominion University, Norfolk, VA 23508, USA 2Instituto de Ciencias Marinas y Limnológicas, Universidad Austral de Chile, Casilla 567, Valdivia 5090000, Chile 3Centro de Investigación Dinámica de Ecosistemas Marinos de Altas Latitudes, Universidad Austral de Chile, Valdivia 5090000, Chile 4COPAS Sur-Austral, Universidad de Concepción, Concepción 4030000, Chile

ABSTRACT: Antarctic krill Euphausia superba, crystal krill E. crystallorophias, and Antarctic sil- verfish Pleuragramma antarctica are key mid-trophic level species in the Ross Sea, connecting pri- mary production to the upper trophic levels. Distributions of these species were constructed from observations made in the western Ross Sea from 1988 to 2004. Distributions of environmental con- ditions were obtained from a 5-km resolution circulation model (temperature, mixed layer depth, surface speed) and satellite-derived observations (chlorophyll, sea ice cover). A hierarchy of sta- tistical methods determined correlations and relationships between species and environmental conditions. Each species occupies a localized habitat defined by different environmental charac- teristics. Antarctic krill are concentrated along the northwestern shelf break in a habitat charac- terized by deep (>1000 m) bottom depth, warm temperature (1 to 1.25°C), decreased sea ice, and proximity to the shelf break. Crystal krill and Antarctic silverfish are concentrated in Terra Nova Bay. Common characteristics of the habitat for these species are southwesterly location, coastal proximity, and cold temperature (−1.75 to −2°C). The habitat characteristics obtained for the 3 spe- cies provide a basis for projecting potential distribution changes in response to environmental change and for delineating regions of the Ross Sea for focused management and selection of marine protected areas that support ecosystem-level conservation plans.

KEY WORDS: Antarctic krill · Euphausia superba · Crystal krill · E. crystallorophias · Pleuragramma antarctica · Habitat characteristics · Ross Sea · Antarctic

INTRODUCTION apex predators (Pinkerton et al. 2010, Ballard et al. 2012, Smith et al. 2012). Two species of krill (Ant arctic The Ross Sea is among the most biologically pro- Euphausia superba and crystal E. crystalloro phias) ductive regions within the Southern Ocean (Arrigo et and one fish species (Antarctic silverfish Pleura- al. 1998b, 2008), contributing about 35% (23.4 ± 9.98 gramma antarctica) account for the majority of the Tg C yr −1 [mean ± SD]) of the annual Southern Ocean mid-trophic level biomass in the Ross Sea and provide shelf production (66.1 Tg C yr−1; Arrigo et al. 2008). an important link between primary production and The primary production of the Ross Sea supports a upper trophic levels (Daly & Macaulay 1988, Ainley et complex and diverse food web that includes a suite of al. 2006, Smith et al. 2007, Pinkerton et al. 2010).

© The authors 2017. Open Access under Creative Commons by *Corresponding author: [email protected] Attribution Licence. Use, distribution and reproduction are un - restricted. Authors and original publication must be credited. Publisher: Inter-Research · www.int-res.com 46 Mar Ecol Prog Ser 584: 45–65, 2017

Many of the apex predators (e.g. penguins, seals, are or are not randomly distributed. The second and cetaceans), fish, and other mid-trophic level spe- objective was addressed by using an empirical cies in the Ross Sea depend directly or indirectly on orthogonal function analysis to establish statistically krill and Antarctic silverfish as their primary food based quantitative relationships between species source (Pinkerton et al. 2010, Smith et al. 2012, and environmental variables that are associated with Pinkerton & Bradford-Grieve 2014). This dependency the krill and silverfish habitat. Addressing these on a few mid-trophic level species sets up a food web objectives provides guidance for understanding the structure in the Ross Sea in which these few species effects of changes in mid-trophic level species on link and control energy flow between low and high ecosystem processes and biogeochemical cycles in trophic levels (Ainley et al. 2015). As a result, changes the Ross Sea. in distribution, abundance, and relative availability of these 3 species affect all trophic levels. Crystal krill and Antarctic silverfish are associated METHODS with sea ice, which provides food and a nursery region for eggs and early life stages (Laws 1985, Kellermann Species’ data sources 1986b, Daly & Macaulay 1988), and these ice-oblig- ate species tend to be found in neritic, ice-covered The Italian Antarctic expeditions that took place in regions (Hubold 1984, Kellermann 1986a, Thomas & the western Ross Sea (Fig. 1A) between 1988 and Green 1988, Hosie 1991, Pakhomov & Perissinotto 2000 provided extensive surveys of the distribution of 1996). For Antarctic krill, the presence of warm Cir- Antarctic krill and crystal krill (Table 1, Fig. 1B). Krill cumpolar Deep Water (CDW) at depth is essential for sampling for juveniles and adults occurred during completion of their early life history (Hofmann et al. the austral summer (December to February) and late 1992, Hofmann & Hüsrevo˘glu 2003), and this species austral spring (November) and extended from 64° to tends to be found in regions where CDW is present 77.5° S and from 177° W to 165° E (Fig. 1B). Antarctic (Quetin & Ross 1984, Azzali & Kalinowski 2000, Sala silverfish surveys were made between 1988 to 2004 et al. 2002, Ashjian et al. 2004, Lawson et al. 2004). as part of cruises that extended from 67° to 78° S and These general distributions exhibit considerable from 175° W to 164° E during the austral late spring spatial variability between seasons and across the and summer (Table 2, Fig. 1C). The abundance data Southern Ocean (Azzali & Kalinowski 2000, Siegel for Antarctic silverfish include the larval, postlarval, 2005, Lawson et al. 2008, Murphy et al. 2017). Studies and juvenile early life history stages. The aggregated have considered relationships be tween krill abun- krill biomass and silverfish abundance distributions dance and different environmental variables such as reported from the cruises listed in Tables 1 & 2, re- temperature, oxygen concentration, sea ice, and salin- spectively, were digitized and converted to common ity (e.g. Witek et al. 1981, Weber et al. 1986, Siegel units of biomass (t km−2) and abundance (ind. 2005, Atkinson et al. 2008, Murphy et al. 2017). How- [1000 m3]–1), respectively. Details of how the biomass ever, predictable relationships that are stable between and abundance distributions were constructed from areas and years have yet to be determined, and the the acoustic and net-based observations are pro- extent to which the distributions of this organism are vided in the references listed in Tables 1 & 2. related to environmental conditions is unknown. Understanding the factors influencing current krill and Antarctic silverfish distributions will allow pro- Environmental data sources jections of distributional changes in response to changing environmental conditions in the Ross Sea The average maximum temperature below 300 m (e.g. Smith et al. 2014b). for the summer season of 2010 (Table 3, Fig. 2A) was The objectives of this study were to (1) construct extracted from simulated circulation distributions ob- distributions of the 3 key mid-trophic level species in tained from a 5-km horizontal resolution implemen- the western Ross Sea and (2) evaluate the extent to tation of the Regional Ocean Modeling System which the distributions of these species are corre- (ROMS) for the Ross Sea (Mack et al. 2017). The lated with environmental variables. The first objec- 300-m temperature distribution provides a tracer for tive was addressed by constructing distributions of CDW which does not exhibit seasonal variations Antarctic krill, crystal krill, and Antarctic silverfish (Dinniman et al. 2003). from historical literature. An optimized hot spot ana - Ocean color measurements (Table 3) from the Sea- lysis was used to test the hypothesis that the species viewing Wide Field-of-view Sensor (SeaWiFS) from Davis et al.: Krill, silverfish, and environmental correlations 47

A Bathymetry (m) Study >1000 Area 500–1000 <500 Land or ice

170°E 180° 170°E 180° B C

70° 70° S S 70° 70° S S A A Cape Cape

I I

D

D Adare

Adare R

R

N N O

O

Krill A Antarctic silverfish A T

T

L L C C Sampling by year Sampling by year

I I

V V 1988 1988 Terra Terra 75° Nova 75° 1989–1990 Nova 1989–1990 Bay Bay 75° 1994 75° 1994 1997 1996 2000 1997, 2000, Ross Ross & 2004 McMurdo Ice Shelllf McMurdo Ice Shelllf N Statiiion Statiiion

150°E 160° 170° 180° 170°W 150°E 160° 170° 180° 170°W Fig. 1. (A) Antarctica, showing the study region in the western Ross Sea and maps of the sampling distributions by year for (B) krill (Euphausia superba and E. crystallorophias) and (C) Antarctic silverfish Pleuragramma antarctica. Bottom bathymetry data are from NOAA National Centers for Environmental Information; International Bathymetric Chart of the Southern Ocean; General Bathymetric Chart of the Oceans

September 1997 to April 2002 were obtained as part construct an average growing season (December to of the Processes Controlling Iron Regulation at the February) climatology (Fig. 2B). Mesoscale (PRISM) project that focused on the Ross Mixed layer depth (MLD; Table 3) was extracted Sea (McGillicuddy et al. 2017). Chlorophyll concen- from simulated circulation distributions for September trations (mg m−3) estimated from SeaWiFS 8 d Level 3 2010 to February 2012 (Mack et al. 2017). The MLD,

binned seasonal climatology files were used to defined as the depth where density (σt) differs from

Table 1. Summary of cruise dates, species sampled, life history stage, units, and reference for the Italian Antarctic Expeditions to the western Ross Sea that provided the krill biomass data sets used in this study. The data sources were screened to avoid duplicate reporting of biomass data sets. nm: nautical mile

Cruise date Species sampled Life history stage Units Reference

January−February 1988 Euphausia crystallorophias Juvenile & adult ind. m−2 Guglielmo et al. (2009) December 1989 − January 1990 E. superba Adult t nm−2 Azzali & Kalinowski (2000) November 1994 E. superba Adult t nm−2 Azzali & Kalinowski (2000) December 1994 E. superba & E. crystallorophias Adult t nm−2 Azzali & Kalinowski (2000) November 1994 E. superba & E. crystallorophias Adult t km−2 Azzali et al. (2006) December 1994 E. superba & E. crystallorophias Adult t km−2 Azzali et al. (2006) December 1997 E. superba & E. crystallorophias Juvenile & adult t km−2 Azzali et al. (2006) January 2000 E. superba & E. crystallorophias Juvenile & adult t km−2 Azzali et al. (2006) January−February 2000 E. superba & E. crystallorophias Juvenile & adult ind. (1000 m3)−1 Sala et al. (2002) 48 Mar Ecol Prog Ser 584: 45–65, 2017

Table 2. Summary of cruise dates, life history stage, units, and references for the Antarctic silverfish data used in this study. The data sources were screened to avoid duplicate reporting of biomass data sets

Life history stage Units Reference

January−February 1988 Postlarvae ind. (100 m3)−1 Guglielmo et al. (1997) November 1989 − January 1990 Postlarvae ind. (100 m3)−1 Granata et al. (2000) November−December 1994 Postlarvae & juvenile ind. (100 m3)−1 Vacchi et al. (1999), Granata et al. (2000) January−February 1996 Juvenile ind. (100 m3)−1 Granata et al. (2000) December 1997 Larvae, postlarvae, & juvenile ind. (1000 m3)−1 La Mesa et al. (2010) January−February 2000 Larvae, postlarvae, & juvenile ind. (1000 m3)−1 La Mesa et al. (2010) January 2004 Larvae, postlarvae, & juvenile ind. (1000 m3)−1 La Mesa et al. (2010)

Table 3. Summary of environmental data, source, data time range, and references used in this study. Environmental dis - tributions were produced from the summer average data (December−February). SeaWiFS: Sea-viewing Wide-Field-of-view Sensor, SSM/I: Special Sensor Microwave Imager. The percent sea ice cover was calculated from the SSM/I measurements using the methodology given in Cavalieri et al. (1996)

Environmental variable Source Time Reference

Maximum temperature Ross Sea Circulation Model December 2010 − March 2011 Mack et al. (2017) below 300 m (°C) Chlorophyll (mg m−3) Ocean Color SeaWiFS September 1997 − April 2002 McGillicuddy et al. (2017) Mixed layer depth (m) Ross Sea Circulation Model September 2010 − February 2012 Mack et al. (2017) Sea ice (% coverage) SSM/I 1988−2003 Cavalieri et al. (1996) Surface speed (m s−1) Ross Sea Circulation Model September 2010 − February 2012 Mack et al. (2017)

the surface by >0.01 kg m−3, was calculated from the tion system package ArcGIS 10.2.2 to construct dis- simulated daily averaged temperature and salinity tributions of the individual variables and to overlay values, and averaged to provide monthly means. The combinations of variables (e.g. bathymetry and krill). monthly MLD values were then averaged to obtain a The distance from each krill and silverfish measure- summer seasonal MLD climatology (Fig. 2C). ment to the location of the coastline and ice shelf The summer average sea ice concentration (Table 3) front, defined by the 0-m isobath, and shelf break, distribution in the Ross Sea was obtained from Spe- defined by the 700-m isobath, was measured using cial Sensor Microwave Imager (SSM/I) measure- the ArcGIS function ‘Near.’ These variables were in - ments made between 1988 and 2003 as part of an cluded as part of the environmental variable data set. analysis of sea ice changes in the Ross Sea (Fig. 3A, Values of the environmental variables that correspond Smith et al. 2014b). to the location of each krill and Antarctic silverfish Estimates of the magnitude and variability of the measurement were extracted using the ArcGIS func- east−west and north−south components of the veloc- tion ‘Extract Multi Values to Points.’ ity field are provided by simulated surface speed distributions that were obtained from the Ross Sea circulation model (Table 3, Mack et al. 2017). The Statistical analyses simulated daily surface speeds were averaged to obtain monthly (December−February) distributions, The optimized hot spot analysis in the spatial statis- which were then averaged to provide a summer dis- tics extension of ArcGIS was used to identify statisti- tribution (Fig. 3B). cally significant spatial clustering of high values (hot spots) and low values (cold spots) in the krill, Antarc- tic silverfish, and environmental distributions using Construction of species distributions and the Getis-Ord Gi* statistic (Getis & Ord 1992). The environmental data sets significance of the Gi* statistic for each data value provides an associated z-score, p-value, and confi- The krill, Antarctic silverfish, and environmental dence level bin. The Gi* value for each data point is data sets were imported into the geographic informa- then partitioned into a bin that reflects its signifi- Davis et al.: Krill, silverfish, and environmental correlations 49

Fig. 2. (A) Simulated summer (December−February) maxi- mum potential temperature (°C) below 300 m obtained from the Ross Sea circulation model (Mack et al. 2017). (B) Aver- age growing season (December−February) chlorophyll (mg m−3) climatology derived from SeaWiFS measurements made in 1997−2002 (McGillicuddy et al. 2017). (C) Summer average (December−February) mixed layer depth (m) distri- bution obtained from simulated temperature and salinity distributions (Mack et al. 2017) 50 Mar Ecol Prog Ser 584: 45–65, 2017

Fig. 3. Summer (December−February) averaged (A) SSM/I-derived distribution of sea ice concentration (0−100% coverage) obtained from measurements made between 1988 and 2003 (Cavalieri et al. 1996), and (B) surface speed (m s−1) obtained from the simulated circulation distributions (Mack et al. 2017) cance level, i.e. confidence intervals of 99% (±3 Gi* no clustering of the data (high values are near low value), 95% (±2 Gi* value), and 90% (±1 Gi* value). values). Very high or low z-scores associated with The remaining Gi* values are considered to be statis- significant p-values reflect a non-random pattern in tically insignificant. the data. The spatial autocorrelation analysis was Histograms that partitioned the number of krill and done for each species and each environmental vari- Antarctic silverfish measurements into 0.25°C tem- able using the entire data set (western Ross Sea perature and 100-m bathymetry bins were used to scale) and using the portion of the data set identified identify initial relationships in the data sets. The by the 99% confidence interval in the hot spot analy- numerous 0 values in the data sets for the 3 species sis (hot spot scale) (Tables S1 & S2 in the Supple- (i.e. absence of the species) biased the histogram fre- ment at www.int-res.com/articles/suppl/ m584 p045_ quencies and obscured underlying patterns. There- supp.pdf ). fore, histograms were constructed that excluded 0 A traditional empirical orthogonal function (EOF) values to better highlight relationships with temper- analysis separates time series measurements at a ature and bathymetry. series of locations into sets of related spatial patterns The spatial autocorrelation (Moran’s I) function with certain time histories (Preisendorfer & Mobley from the spatial statistics extension of ArcGIS was 1988). The EOF implementation used in this study used to measure the similarity of features. The analyzes a set of variables, i.e. the species and envi- Moran’s I statistic uses the data locations and values ronmental distributions, at a series of locations. The to estimate the degree of similarity in the data and EOF yields patterns of related variables, given by the the type of similarity, i.e. clustered, dispersed, or ran- eigenfunction, and locations at which the patterns dom. A positive Moran’s I index indicates that high are dominant, the loading vector. The eigenvalue for values cluster near high values and low values clus- each mode in dicates the fraction of the variance of ter near low values, signifying spatially clustered dis- the data set represented by a given mode. Significant tributions, and a negative Moran’s I index indicates modes are identified by those with eigenvalues that Davis et al.: Krill, silverfish, and environmental correlations 51

are greater than the mean of all the eigenvalues, i.e. in offshelf waters (Fig. 4C). Localized regions of en- ‘the average eigenvalue test’ (Preisendorfer & Mob- hanced abundance occur on the shelf in the shal- ley 1988). The eigenfunctions indicate which vari- lower waters over Mawson and Pennell Banks. The ables are related for a given mode, and the loading largest and most consistent occurrence of Antarctic vector indicates the dominance of a pattern. The silverfish is in Terra Nova Bay (Fig. 4C), where abun- eigenvector provides the direction of change (posi- dance values exceed 10 000 ind. (1000 m3)−1. Just out- tive or negative) relative to the mean. The EOF side of Terra Nova Bay, along the southwestern flank analysis was done for the species and environmental of Mawson Bank, Antarctic silverfish abundance lev- data in the region defined by the chlorophyll distri- els are similar to those in Terra Nova Bay (Fig. 4C). bution (Fig. 2B), which sets a common area for all The histograms show that the largest number of data sets. The analysis was done using the entire Antarctic krill observations are associated with tem- data set (western Ross Sea scale) and the data set peratures of 1 to 1.25°C and bottom depths of 1000 m defined by the 99% confidence interval obtained for (Fig. 5A). The Antarctic krill histogram shows sec- the biological hot spot (hot spot scale). The mean ondary peaks at colder temperatures (−1.25 to −1.5°C) value was subtracted from each data set to remove and shallower bottom depths (300 and 400 m). The the effects of trends so that the data variability is bet- largest number of crystal krill observations are asso- ter represented. The resultant data set was divided ciated with temperatures of −1.75 to −2°C and bottom by the standard deviation of the data to create a 0- depths between 500 and 800 m (Fig. 5B). Secondary mean unit variance data set. This detrended and peaks in the crystal krill histogram are at −1.0 to scaled data set was used in the EOF analysis. −1.5°C and 300, 400, and 500 m. The temperatures and bottom depths that correspond to the largest number of observations of the early life stages of RESULTS Antarctic silverfish are similar to those associated with crystal krill, −1.75 to −2°C and 500 and 700 m Distributions (Fig. 5C). The fewest observations occur at warm temperatures and 900 m. Antarctic krill are found over most of the western Ross Sea continental shelf, along the shelf break, and in offshelf waters (Fig. 4A). The inner shelf region is Hot spot and spatial autocorrelation analysis characterized by low biomass (<1 t km−2) or absence of Antarctic krill (Fig. 4A). The exception is the The optimized hot spot analysis shows that the southwest inner shelf, where Antarctic krill biomass Antarctic krill distribution over most of the Ross Sea was consistently 1−10 t km−2. The shallow areas over continental shelf is not significantly different from an Mawson and Pennell Banks have the largest biomass average distribution (Fig. 6A). The exception is the of 100−1000 t km−2. Antarctic krill biomass is consis- northwestern shelf/slope region. The area defined tently 100−1000 t km−2 along the shelf break/slope by the 99% confidence level (red circles) outlines a region and in the northwestern corner of the Ross biological hot spot (region enhanced relative to sur- Sea (Fig. 4A). Biomass values in excess of 1000 t km−2 rounding areas) for Antarctic krill along and off the occur in the northwest Ross Sea off the coast of Cape shelf/slope region to the east of Cape Adare (Fig. 6A). Adare (Fig. 4A). The 90% confidence level expands this region to the Crystal krill occur at low biomass levels (less than 1 t north and southward onto the outer continental shelf km−2) over most of the western Ross Sea continental (Fig. 6A). The crystal krill distribution is relatively shelf and were not found in samples from the eastern uniform over most of the continental shelf, except in part of the study region (Fig. 4B). Localized areas of Terra Nova Bay (Fig. 6B), where a distinct hot spot biomass of 1−10 t km−2 occur over Mawson and Pen- occurs. The region outlined by the 99% confidence nell Banks, the southern inner shelf near the Ross level includes the inner and northern shelf region of Ice Shelf, and the southwestern inner shelf (Fig. 4B). Terra Nova Bay and the outer part of the Drygalski Biomass of 100−1000 t km−2, the highest observed, Ice Tongue. The region encompassed by the 90% occurs in and around Terra Nova Bay in the far west- con fidence interval slightly expands the hot spot area ern Ross Sea. (Fig. 6B). The biological hot spot of the early life The early life stages of Antarctic silverfish occur at stages of Antarctic silverfish is also con centrated in low (<10 ind. [1000 m3]−1) abundance levels over Terra Nova Bay and in the near environs with a dis - most of the western continental shelf and are absent tribution that is similar to crystal krill (Fig. 6C). 52 Mar Ecol Prog Ser 584: 45–65, 2017

Fig. 4. Biomass distributions of (A) Antarctic krill and (B) crystal krill constructed from the data sources in Table 1 ex- pressed as t km−2. (C) Abundance distribution of Antarctic silverfish early life history stages constructed from the data sources in Table 2 expressed as ind. (1000 m3)−1. Bottom ba- thymetry is given in m. Mawson Bank (MB) and Pennell Bank (PB) are indicated Davis et al.: Krill, silverfish, and environmental correlations 53

70 verfish values are clustered at a statistically signifi- A 1.75 60 cant level (p < 0.01, Table S1). 1.25 At the hot spot scale, the Antarctic and crystal krill 50 0.75 biomass distributions are statistically random (z-

40 0.25 score = −0.1272, p = 0.8988 and z-score = −0.6648, p = 0.5062, respectively, Table S2). The Antarctic silver- 30 -0.25 fish abundance distribution is statistically clustered

20 -0.75 at the hot spot scale, as are the associated environ- -1.25 mental variables (z-score > 2.58, p < 0.01, Table S2). 10 -1.75 0 200 300 400 500 600 700 800 900 1000 EOF analysis 25 B 1.75 Western Ross Sea scale 20 1.25 0.75 The eigenvalues obtained from the EOF analysis 15 0.25 (Table S3) show that the first 4 modes are significant for Antarctic krill and crystal krill and the first -0.25 10 3 modes are significant for the Antarctic silverfish -0.75 (Fig. 7) at the western Ross Sea scale (Table 4).

No. of observations 5 -1.25 The first mode obtained for the Antarctic krill west- Potential temperature (°C) -1.75 ern Ross Sea scale data set (Fig. 8A, blue bars) ex - 0 plains 38.04% of the variance (Table S3) and is ac- 200 300 400 500600 700 800 900 1000 20 counted for by the association of southerly (northerly) 1.75 latitudes, cold (warm) water temperature, higher 18 C (lower) chlorophyll concentrations, and in creased 16 1.25 (decreased) distance to the shelf break (Table 4). 14 0.75 Mode 2 (Fig. 8A, cyan bars) explains 16.78% of the 12 0.25 variability in the data set (Table S3), and the eigen- 10 vectors indicate that this is attributed to bottom -0.25 8 depth, sea ice, and surface speed. Mode 2 is associ- 6 -0.75 ated with higher concentrations of sea ice, slower 4 -1.25 surface speeds, and deeper bottom depths (Table 4). 2 Modes 3 and 4 (Fig. 8A, yellow and red bars) account -1.75 0 for 11.86% and 10.49% (Table S3) of the variance, 200 300 400 500600 700 800 900 1000 respectively, and suggest a correspondence between Depth (m) longitude and MLD for Mode 3, and longitude, sur- Fig. 5. Frequency distribution of (A) Antarctic krill, (B) crys- face speed, and distance to the coast for Mode 4 tal krill, and (C) Antarctic silverfish observations (excluding (Table 4). 0 values) as a function of potential temperature (°C) and bot- tom depth (m). Chi-squared goodness-of-fit tests indicate The 4 significant modes obtained for the crystal that the 3 histograms differ from a normal distribution, with krill data set at the western Ross Sea Scale explain p-values of <0.0005, 0.0022, and <0.0005, respectively 36.31, 17.53, 10.6, and 9.21% of the variance (Table S3), respectively. Mode 1 is associated with latitude, distance to the shelf break, distance to the coast, and The spatial autocorrelation analysis indicates that temperature (Fig. 8B, blue bars), suggesting corre- the distribution pattern of Antarctic krill is statisti- spondences between northerly (southerly) latitudes, cally random at the western Ross Sea scale (z-score = increased (decreased) distance from the coast, de - 0.7597, p = 0.4474, Table S1 in the Supplement). In creased (increased) distance to the shelf break, and contrast, the distributions of adult crystal krill and warmer (colder) temperatures (Table 4). The vari- early life stages of Antarctic silverfish are clustered ance in Mode 2 is attributed to bottom depth, chloro- at a statistically significant level at the western Ross phyll, and sea ice (Fig. 8B, cyan bars), suggesting an Sea scale (p < 0.01, Table S1). The environmental association among shallow bottom depths, decreased variables associated with the krill and Antarctic sil- concentrations of sea ice, and higher concentrations 54 Mar Ecol Prog Ser 584: 45–65, 2017

Fig. 6. Distribution of regions with clustering that is signifi- cant at the 99, 95, and 90% significance level for (A) Ant - arctic krill, (B) crystal krill, and (C) Antarctic silverfish early life history stages obtained from the optimized hot spot analysis. Regions that do not have significant clustering are also shown Davis et al.: Krill, silverfish, and environmental correlations 55

4.5 is associated with eastern longitudes, distance to the 4 Antarctic krill shelf break, distance to the coast, and temperature Crystal krill (Fig. 8C, blue bars), suggesting that eastern (west- Antarctic silverfish 3.5 ern) longitudes are associated with warmer (colder) 3 temperatures, decreased (increased) distance from the shelf break, and increased (decreased) distance 2.5 to the coast (Table 4). The variance in Mode 2 is

2 accounted for by bottom depth, sea ice, and chloro- phyll concentration (Table 4), suggesting an associa- Eigenvalue 1.5 tion among deeper bottom depths, increased sea ice

1 coverage, and decreased chlorophyll concentrations (Fig. 8C, cyan bars). Mode 3 provides an associa- 0.5 tion among bottom depth, surface speed, and MLD

0 (Table 4), in which slower surface speeds are associ- 1234567891011ated with deeper bottom depths and deeper mixed Mode layers (Fig. 8C, yellow bars). Fig. 7. Eigenvalues and associated modes for Antarctic krill, crystal krill, and Antarctic silverfish obtained from the entire data set (western Ross Sea scale). Modes with eigenvalues greater than the mean (=1, dashed line) are significant Hot spot scale

The eigenvalues obtained from the EOF analysis of chlorophyll. Modes 3 and 4 indicate correspon- for each species and associated environmental distri- dences between longitude and MLD (Fig. 8B, yellow butions at their associated hot spot scale (Table S4) bars), and crystal krill biomass, bottom depth, and show that the first 4 modes are significant for Antarc- MLD (Fig. 8B, red bars), respectively (Table 4). Mode tic krill and Antarctic silverfish, and the first 3 modes 4 indicates that decreased (increased) crystal krill are significant for crystal krill (Fig. 9). biomass is associated with increased (decreased) bot- The 4 significant modes for the Antarctic krill data tom depths and deeper (shallower) mixed layers set at their associated hot spot explain 43.93, 23.3, (Table 4). 12.86, and 10% of the variance (Table S4), re - The 3 significant modes obtained for Antarctic sil- spectively. Mode 1 characterizes their northwestern verfish explain 39.74, 21.5, and 12.59% of the vari- corner hot spot with bottom depth and water tempera- ance in the data set (Table S3), respectively. Mode 1 ture (Fig. 10A, blue bars). Mode 2 further refines their

Table 4. Summary of the dominant variables and direction of change (positive: +, negative: −) relative to the mean for each species at the western Ross Sea scale and the hot spot scale obtained from the EOF analysis. Long: longitude, Lat: latitude, Bio: biomass, Abu: abundance, Depth: bottom depth, Speed: surface speed, Temp: temperature below 300 m, Coast: distance to coastline, Ice: sea ice coverage, Chla: chlorophyll a concentration, Shelf: distance to shelf break, MLD: mixed layer depth

Western Ross Sea scale Hot spot scale Significant mode Dominant variables Significant mode Dominant variables

Antarctic 1 Lat (−), Temp (−), Chla (+), Shelf (+) 1 Depth (+), Temp (+) krill 2 Depth (+), Speed (−), Ice (+) 2 Ice (−), Shelf (−) 3 Long (−), Ice (+), MLD (−) 3 Lat (+), Speed (+) 4 Long (−), Speed (−), Coast (+) 4 Bio (−), MLD (−) Crystal 1 Lat (+), Temp (+), Coast (+), 1 Long (−), Temp (−), Coast (−), krill Shelf (−) Chla (−), Shelf (+) 2 Depth (−), Ice (−), Chla (+) 2 Lat (−), Depth (+), MLD (−) 3 Long (−), MLD (−) 3 Lat (+), Bio (+), Temp (−) 4 Bio (−), Depth (+), MLD (+) Antarctic 1 Long (+), Temp (+), Coast (+), 1 Long (−), Speed (−), silverfish Shelf (−) Temp (−), Coast (−) 2 Depth (+), Ice (+), Chla (−) 2 Depth (−), Ice (+), MLD (+) 3 Depth (+), Speed (−), MLD (+) 3 Lat (−) 4 Abu (+), Chla (−) 56 Mar Ecol Prog Ser 584: 45–65, 2017

A Mode 1 0.5 Mode 2 Mode 3 0 Mode 4 Eigenvector

Antarctic krill -0.5

Long LatBio Depth Speed Temp Coast Ice Chla Shelf MLD

B Mode 1 0.5 Mode 2 Mode 3 0 Mode 4 Crystal krill Eigenvector -0.5

Long LatBio Depth Speed Temp Coast Ice Chla Shelf MLD

C Mode 1 0.5 Mode 2 Mode 3 0

Eigenvector -0.5 Antarctic silverfish Long LatAbu Depth Speed Temp Coast Ice Chla Shelf MLD Environmental variables Fig. 8. Eigenvectors showing partitioning of variance for each significant EOF mode and environmental variable obtained from the entire data set (western Ross Sea scale) for (A) Antarctic krill, (B) crystal krill, and (C) Antarctic silverfish. Long: lon- gitude, Lat: latitude, Bio: biomass, Abu: abundance, Depth: bottom depth, Speed: surface speed, Temp: temperature below 300 m, Coast: distance to coastline, Ice: sea ice coverage, Chla: chlorophyll a concentration, Shelf: distance to shelf break, MLD: mixed layer depth habitat by including sea ice cover and distance to 6 the shelf break (Fig. 10A, cyan bars). The eigenvectors for mode 3 indicate an association between surface Antarctic krill speed and latitude (Fig. 10A, yellow bars), and a cor- 5 Crystal krill Antarctic silverfish respondence between Antarctic krill biomass and

MLD is given by Mode 4 (Fig. 10A, red bars). The first 4 3 modes suggest that the hot spot habitat for Antarctic krill in the northwestern corner of the Ross Sea shelf break region is defined by deep bottom depths with 3

warm water temperatures, decreased sea ice, and fast Eigenvalue surface speeds (Table 4). Mode 4 of the EOF suggests 2 that within this habitat, increased Antarctic krill bio- mass is associated with waters that have deeper 1 mixed layers (Table 4). The 3 significant modes at the hot spot scale for crystal krill explain 39.79, 17.86, and 15.02% of the 0 1234567891011 variance in the data set, respectively (Table S4). The Mode eigenvectors for Mode 1 capture the Terra Nova Bay characteristics between longitude, distance to the Fig. 9. Eigenvalues and associated modes for Antarctic krill, crystal krill, and Antarctic silverfish obtained from the data coast, temperature, and chlorophyll concentration set defined by the 99% confidence interval from the hot spot (Fig. 10B, blue bars). Mode 2 further refines the analysis (hot spot scale). Modes with eigenvalues greater habitat characteristics by including bottom depth, than the mean (=1, dashed line) are significant Davis et al.: Krill, silverfish, and environmental correlations 57

A Mode 1 0.5 Mode 2 Mode 3 0 Mode 4 Eigenvector

Antarctic krill -0.5

Long LatBio Depth Speed Temp Coast Ice Chla Shelf MLD

Mode 1 0.5 B Mode 2 Mode 3 0 Crystal krill Eigenvector -0.5

Long LatBio Depth Speed Temp Coast Ice Chla Shelf MLD

1 C Mode 1 Mode 2 Mode 3 0 Mode 4 Eigenvector

Antarctic silverfish -1 Long LatAbu Depth Speed Temp Coast Ice Chla Shelf MLD Environmental variables Fig. 10. Eigenvectors showing partitioning of variance for each significant EOF mode and environmental variable obtained from the data set defined by the 99% confidence level from the hot spot analysis (hot spot scale) for (A) Antarctic krill, (B) crystal krill, and (C) Antarctic silverfish. Environmental variables are abbreviated as in Fig. 8 latitude, and MLD (Fig. 10B, cyan bars). The eigen- tions with slow surface speeds, cold temperatures, vectors for Mode 3 show an association among crys- shallow bottom depths, deep MLDs, and increased tal krill biomass, water temperature, and latitude sea ice cover (Fig. 10C). Within the Terra Nova Bay (Table 4). The first 2 modes suggest that the hot spot habitat, increased abundance of the early life stages habitat in Terra Nova Bay for crystal krill is defined of Antarctic silverfish is associated with low chloro- by westerly longitudes with cold temperatures, low phyll concentrations (Table 4). chlorophyll concentrations, and deep bottom depths. Mode 3 of the EOF suggests that within Terra Nova Bay, increased crystal krill biomass is associated with DISCUSSION the northern latitudes within this habitat and cold temperatures (Fig. 10B, yellow bars). Perspective The 4 significant modes from the EOF analysis at the Terra Nova Bay hot spot scale for Antarctic silver- The relationships developed to describe correspon- fish explain 47.88, 21.66, 10.95, and 10.26%, respec- dences between terrestrial and marine species and tively (Table S4). Mode 1 indicates an association environmental conditions take many forms (see between longitude, surface speed, temperature, and review by Elith & Leathwick 2009), but all have the distance to the coast and shelf (Fig. 10C, blue bars). goal of providing quantitative and robust models that Modes 2 and 3 add bottom depth, MLD, ice, and lati- allow the prediction of potential effects of changing tude as additional habitat characteristics, respec- environmental conditions on species distributions. tively (Fig. 10C, cyan and yellow bars). Mode 4 pro- For the Antarctic marine system, species distribution vides a correspondence between Antarctic silverfish models have focused on Antarctic krill and top pred- abundance and chlorophyll concentration (Fig. 10C, ators because of their importance in the food web, red bars). These in total suggest that the hot spot and the availability of observations to describe the habitat for the early life stages of Antarctic silverfish distributions and habitat characteristics (Ichii 1990, in Terra Nova Bay is defined by southwesterly loca- Trathan et al. 2003, Naganobu et al. 2006, Atkinson 58 Mar Ecol Prog Ser 584: 45–65, 2017

et al. 2008, Murase et al. 2013). Early attempts to primary relationships that emerge from this analysis explain the observed circumpolar distribution of are those that characterize the physical environment Antarctic krill were based on qualitative comparisons of the Ross Sea. Observations from all seasons taken with environmental conditions, such as temperature, with space and time resolution that reveals variabil- sea ice, and bathymetry (Marr 1962, Mackintosh ity are needed to define the full range of the habitats 1973). While providing insights into basic factors occupied by the 3 species. affecting Antarctic krill distributions, these qualita- The general characteristics of the distributions of tive studies lacked the ability to quantify relation- the 3 species show clear and distinct patterns in the ships with environmental distributions, and to assess Ross Sea. Each species occupies a localized habitat regional differences in controls. More recent studies that is defined by different environmental character- have developed statistical relationships and models istics (Table 5). The data sources to estimate the krill that allow quantification of the relative effects of biomass and Antarctic silverfish abundance used in environmental conditions on circumpolar (Hofmann the EOF analysis are based on acoustic measure- & Hüsrevo˘glu 2003, Atkinson et al. 2004, 2008) and ments and net tows (e.g. oblique tows), respectively, regional distributions of Antarctic krill (Weber & El- which integrate over the upper 120 to 160 m. Thus, Sayed 1985, Witek et al. 1988, Trathan et al. 2003, the habitat characteristics are representative of the Lawson et al. 2008). These studies identified patterns upper water column rather than specific depths where and provided correlations between Antarctic krill the krill and Antarctic silverfish were collected. and environmental variables, but also showed that Antarctic krill occupy the habitat along the outer these exhibited considerable variability, which is in continental shelf break. Crystal krill and the early life part attributed to different space and time scales that stages of Antarctic silverfish occupy similar and over- are resolved by the observations used in the analy- lapping habitats in the Terra Nova Bay region that ses. Similar studies for other krill species, such as are distinct from the Antarctic krill habitat. Separa- crystal krill, have not been done, except through tion of Antarctic krill and crystal krill habitats in the qualitative analyses that relate presence to ambient western Ross Sea has been described for specific conditions (e.g. Fevolden 1980, Brinton & Townsend times and locations (Sala et al. 2002, Taki et al. 2008, 1991, Daly & Zimmerman 2004). The general distri- Leonori et al. 2017), but this study shows that the sep- bution of the early life history stages of Antarctic sil- aration is a general characteristic of the 2 species. verfish is known (see review by La Mesa & Eastman Interactions between crystal krill and Antarctic sil- 2012), but again correspondences with environmen- verfish in the Terra Nova Bay habitat are not seen tal conditions are based on qualitative analyses (Hu - from the EOF analysis, but biological interactions bold 1984, Kellermann 1986a,b, Koubbi et al. 2011). such as predator−prey responses are not captured in Our analysis advances these studies by quantita- the data used in this analysis. Some overlap in the tively characterizing the habitats of 3 important mid- distribution of the 3 species occurs over the continen- trophic level species in the Ross Sea. The use of tal shelf, but the regions where the biomass is en- environmental distributions obtained from a high- hanced (i.e. statistically significant) relative to aver- resolution circulation model provides a range of envi- age conditions, i.e. the hot spots for the 2 krill species, ronmental distributions at the scale of the Ross Sea are separate and distinct. that are not possible to obtain from observations. The addition of satellite-derived distributions provides a comprehensive characterization of the environmen- Habitat characteristics tal conditions, that when combined with ship-based biomass distributions, provide a basis for quantitative Antarctic krill assessments of distributional patterns and habitats of each species. Although Antarctic krill are distributed throughout Observations of the 3 species included in this the Ross Sea, the habitat defined by this analysis indi- analysis come from cruises made in the austral spring cates that this species is found in areas overlying and summer; observations from other times of the warm water at depth with decreased sea ice concen- year are lacking due to the logistical constraints of trations and proximity to the shelf break (Table 5), i.e. making ship-based measurements in the Ross Sea. conditions that are found off and along the northwest- These observations capture general characteristics of ern shelf/slope region. Many studies have shown that summer conditions, but are not sufficiently dense in Antarctic krill are associated with the shelf/slope re- space or time to capture variability. As a result, the gion throughout the Antarctic (Hosie 1991, Miquel Davis et al.: Krill, silverfish, and environmental correlations 59

1991, Trathan et al. 2003, Ashjian et al. 2004, Siegel 2005, Azzali et al. 2006, Taki et al. 2008) and note that this area is important to many parts of their life history (Nicol 2006). Moreover, the circumpolar analysis by Atkinson et al. (2008) found that 87% of the Antarctic krill population occurs over deep water and speculated that this habitat protects against intense predation pressure found in shal- lower shelf regions, where the species is accessible to a range of predators. The Antarctic krill hot spot region in the north-

xed layer Cold temperature western Ross Sea is associated with shelf/slope les of the first 2 EOF modes hot

the biomass and abundance associations waters of >1000 m and temperatures of 1 to 1.25°C

(Table 5). This hot spot overlies an area where m the optimized hot spot analysis; bathym- intrusions of CDW occur along the outer shelf region (Dinniman et al. 2003, Smith et al. 2007, Orsi

No association Low chlorophyll & Wiederwohl 2009), which is considered neces- sary for successful completion of the first phase of on Shallow depth Northerly latitude their early life history (Ross et al. 1988, Hofmann et al. 1992, Hofmann & Hüsrevo˘glu 2003). The north- No association Deep mixed layer western Ross Sea is a spawning region for Antarc-

haracteristics Biomass & abundance associations tic krill (Spiridonov 1996) further supporting iden- tification of this area as a habitat. The slower surface speeds associated with the habitat in the northwestern corner of the Ross Sea provide a mechanism for maintaining this region as a hot spot. Simulated circulation distributions (Din- Western Ross Sea scale Hot spot scale niman et al. 2003) show alternating regions of on− Decreased sea ice Proximity to shelf break Cold temperature Low chlorophyll Deep water Shallow mixed layers Slow surface speed Cold temperature Shallow water Increased sea ice offshelf flow along the northwestern shelf with about a 20−25 km scale that are associated with a closed gyre circulation. Numerical Lagrangian simulations implemented for the western Ross Sea in which particles were released along the shelf/ slope and over the continental shelf show on− offshelf transport along the shelf/slope region (Piñones et al. 2016). Particle transport was to the northwest along the outer shelf with retention in the shelf/ slope region that corresponds to the Antarctic krill hot spot. Estimates of Antarctic krill life history Coastline proximity Shallow mi Coastline proximity stage, based on particle transport times, showed that the aggregated particles corresponded to larval

shelf break Warm temperature and adult stages (Piñones et al. 2016). The Antarctic krill distribution in the hot spot region is statistically random. This suggests that the distribution at this scale is variable and unpre- dictable, similar to results from studies in other Ant arctic regions (Priddle et al. 1988, Sushin & Shulgovsky 1999, Trathan et al. 2003, Lawson et al. 2008). Krill are not only structured by environmen- tal conditions (such as currents), but also by behav- ioral and ecological pressures, such as the acquisi- tion of food and avoidance of predators, which E. crystallorophias Euphausia superba Pleuragramma antarctica location

Hot spot Bathymetry Temperature Habitat c Crystal krill Nova Bay Terra 500−800 m −1.75 to −2°C Southwesterly locati Antarctic silverfish Nova Bay Terra 500−700 m −1.75 to −2°C Southwesterly location Antarctic krill Northwest >1000 m 1 to 1.25°C Deep water Table 5. Summary of hot spot locations and habitat characteristics for each species. The wereTable determined fro etry and temperature ranges areetry from the histogram analyses; and habitat characteristics are defined by the dominant variab spot analysis, which explained more than 50% of the data variance. The EOF analysis at Ross Sea and hot spot scales provide affect variability and regional distributions (Daly & 60 Mar Ecol Prog Ser 584: 45–65, 2017

Macaulay 1991); these factors were not included in complete description of habitat characteristics is pro- this analysis. The EOF analysis also indicated that vided by including in situ chlorophyll measurements enhanced Antarctic krill biomass is associated with that are coincident with crystal krill distributions (e.g. deeper mixed layers, which may be related to food Leonori et al. 2017), which resolve relevant scales. supply, although no correspondence with surface Although speed did not emerge as a primary habi- chlorophyll emerged from the analysis. tat characteristic at either scale, results from numeri- cal Lagrangian simulations show particle retention in Terra Nova Bay and over the shallow banks in the Crystal krill western Ross Sea (Piñones et al. 2016). The crystal krill biomass data set used in the EOF analysis had Crystal krill are typically found in inshore neritic too few observations (<1% of total data) from over environments throughout the Antarctic (Fevolden the shallow banks to support the suggestion that 1980, O’Brien 1987, Thomas & Green 1988, Hosie these are retention areas for crystal krill. In Terra 1991, Nordhausen 1994, Pakhomov & Perissinotto Nova Bay, the sampling distribution for crystal krill 1996, Sala et al. 2002, Daly & Zimmerman 2004, covered primarily the continental shelf (Guglielmo et Amakasu et al. 2011). In the Ross Sea, crystal krill are al. 2009), which is an area of uniform slow speeds in widely distributed over the western continental shelf the simulated circulation distribution. As a result, a and are concentrated in Terra Nova Bay. The general relationship with speed cannot be determined be - habitat characteristics for crystal krill determined cause the biomass values are associated with the same from this study consist of southwesterly locations speed values. Enhanced spatial resolution in the near the coast with cold temperatures (1.75 to −2°C), sampling of crystal krill distributions in Terra Nova low chlorophyll, bottom depths between 500 and Bay and over the shallow banks is needed to resolve 800 m, and shallow mixed layers (Table 5). These and better define habitat relationships for this region. conditions define a primary habitat in Terra Nova Bay, which supports a statistically significant region of enhanced crystal krill biomass. Antarctic silverfish The crystal krill distribution at the western Ross Sea scale is clustered at a statistically significant level, Antarctic silverfish are typically found over conti- whereas the distribution in Terra Nova Bay is statisti- nental shelf waters throughout the Antarctic (Hubold cally random. This suggests that gradients in environ- 1984, DeWitt et al. 1990, White & Piatkowski 1993, mental variables are structuring the crystal krill at Trunov 2001, Donnelly & Torres 2008) and exhibit a larger scales, but their distributions at smaller scales circum-Antarctic distribution (Emslie & McDaniel are responding to ecological and behavioral controls. 2002). This species has a wide distribution through- The EOF analysis at the western Ross Sea scale out the Ross Sea continental shelf (DeWitt 1970, showed that increased crystal krill biomass is associ- Granata et al. 2000, Ackley et al. 2003, Donnelly et al. ated with shallow bathymetry and shallow mixed lay- 2004, O’Driscoll et al. 2011), but occurs in higher ers, i.e. conditions that are associated with the shallow abundance in particular regions. The habitat defined banks that extend across the continental shelf (Smith by this analysis indicates that the early life history et al. 2014b). The EOF analysis indicated that in- stages of this species are found in southwesterly creased crystal krill biomass is associated with colder locations with cold temperatures (−1.75 to −2°C), bot- water temperatures and northerly latitudes in Terra tom depths of 500 and 700 m, slow surface speeds, Nova Bay, which is consistent with observations that increased sea ice concentration, deep mixed layers, showed highest crystal krill biomass in the northerly and proximity to the coast (Table 5). Other features region of Terra Nova Bay (Guglielmo et al. 2009). are also important factors for the early life history of The inverse relationship with chlorophyll that silverfish, including proximity to an ice shelf or emerged from the EOF analysis as a habitat charac- glacial tongue, canyons, water mass stratification, teristic for crystal krill is counter to the positive rela- polynyas, and katabatic winds (review in Vacchi et tionship found by Leonori et al. (2017). The satellite- al. 2012b), all of which are features associated with derived chlorophyll climatology used for the EOF Terra Nova Bay. The Antarctic silverfish distributions analysis shows values in Terra Nova Bay that are are clustered at a statistically significant level at the generally <0.5 mg m−3 and does not resolve varia - western Ross Sea and Terra Nova Bay scales, sug- bility at mesoscale or sub-mesoscale scales, which gesting similar environmental structuring of their are likely the scales that crystal krill target. A more distributions. Davis et al.: Krill, silverfish, and environmental correlations 61

The highest abundance of Antarctic silverfish early tolerate only a small range of temperatures) and have life stages occurs in Terra Nova Bay, which was life cycles that are dependent on sea ice, making them identified as a statistically significant hot spot and is potentially sensitive to climatic change. known as a nursey area for silverfish eggs and larvae Summer seasonal sea ice coverage is projected to (Vacchi et al. 2012a). Antarctic silverfish tend to decrease to 44% of current coverage, and that for spawn in ice-covered coastal regions, ice shelf areas, 2100 is projected to decrease further to 22% of cur- and polynyas (Vacchi et al. 2004, 2012a), and use sea rent conditions (Smith et al. 2014b). Habitat observa- ice as a nursery for their eggs and early life stages tions show that crystal krill and Antarctic silverfish (Kellermann 1986a,b, Daneri & Carlini 2002). Antarc- occupy ice-covered regions. Reduction of this habitat tic silverfish spawn in the lower water column, and may impact both species through effects on food sup- embryos, which are slightly positively buoyant, are ply, adult reproduction, and availability of nursery found associated with sea-ice thickness and platelet grounds for early life history stages. Both species ice (Vacchi et al. 2004, 2012a,b). The slow surface have life history strategies that are timed with the speeds that emerged as a characteristic for the Antarc- opening of polynyas, and earlier polynya formation tic silverfish habitat will facilitate retention of the may result in a mismatch with their required envi- early life stages in the productive polynya region. ronmental conditions, producing decreased repro- Unlike the crystal krill, the distribution of Antarctic ductive success and survival. Reduced sea ice may silverfish observations (La Mesa et al. 2010) was suf- favor early spawning by Antarctic krill and increase ficient to capture speed as a habitat characteristic, recruitment (Siegel & Loeb 1995) due to the extended which highlights the importance of designing sam- growth period throughout the summer months. How- pling strategies to include scales of environmental ever, this species also depends on sea ice as an over- controls. wintering habitat that provides food sources and pro- The EOF analysis indicated that sea ice was not a tection for early life history stages (Daly 1990, 2004) strong factor structuring the Terra Nova Bay habitat, as well as older stages (Siegel & Loeb 1995). which is counterintuitive for a species that depends Projected decreases in CDW inputs to the Ross Sea on sea ice for reproduction and food supply. This continental shelf and shallower MLDs will potentially result is likely an artifact of the ship-based sampling, reduce nutrient supply to the upper water column which avoids regions with high sea ice coverage. For and allow the ocean to remain stratified for a longer the Terra Nova Bay hot spot, the EOF analysis indi- period of time (Smith et al. 2014b). Projections for cated that increased Antarctic silverfish abundance 2050 suggest that these changes will reduce the rela- is associated with low chlorophyll concentrations, tive contribution of Phaeocystis antarcticum to the which is also counterintuitive given the high levels continental shelf production and favor diatom pro- of primary production in Terra Nova Bay (Arrigo et duction (Smith et al. 2014a,b, Kaufman et al. 2017). al. 1998a). As for crystal krill, the satellite-derived This may provide increased food supply for all 3 spe- chlorophyll climatology did not resolve variability at cies, especially Antarctic krill, which prefer diatoms scales that are relevant to Antarctic silverfish. as food (Haberman et al. 2003). However, reduced CDW may adversely affect the ability of Antarctic krill to complete the descent−ascent portion of their Climate considerations early life history, which depends on warm water at depth to accelerate embryo and early larval stage The habitat characterizations obtained from this development. study provide a basis for understanding potential The projected changes in environmental conditions changes in the distribution of the 3 species as climate may also alter the present distribution of hot spots, affects environmental conditions in the Ross Sea. Pro- which are foci for locally intense energy fluxes within jections of conditions for 2050 and 2100 that are based the food web (Atkinson et al. 2001, Murphy et al. on assumed warming trends and changes in wind 2007). Changes in environmental conditions that alter forcing suggest that the Ross Sea will experience de- hot spot location, size, and persistence potentially af- creased summer sea ice concentration and extent, fect the distribution, abundance, and relative avail- earlier formation of polynyas, expansion of the Ross ability of the 3 mid-trophic level species, that in turn, Sea polynya, shallower summer MLDs, and a decrease affect all trophic levels of the Ross Sea food web. Ain- in the advection of CDW onto the shelf (Smith et al. ley et al. (2006, 2015) showed that reduced availability 2014b). The 3 species of interest to this study are suc- of the 2 krill species in late austral summer resulted in cessful at present, but they are stenothermal (able to a trophic cascade that caused diet switching, changes 62 Mar Ecol Prog Ser 584: 45–65, 2017

in foraging behavior, and reduced abundance of top Acknowledgements. This study was supported by the Nat- predators. The hot spot locations identified in this ional Science Foundation, Office of Polar Programs Grants ANT-0944174 and ANT-0838911, and Old Dominion Uni- study are in areas associated with land-based colonies versity teaching and research assistantships to L.B.D., and of penguins (Woehler 1993, Ainley et al. 2010, Smith A.P. thanks CONICYT for support through the Program et al. 2014a) and en hanced abundance of seals and FONDAP, Project 15150003. Thoughtful comments provided cetaceans (Stirling 1969, see review by Smith et al. by Peter Sedwick, Mark Butler, and Ari Friedlaender improved an earlier version of this manuscript. Statistical 2007). Hence, alterations to these sites have poten- advice from Chester Grosch, data contributions from Dennis tially large consequences for the Ross Sea food web. McGillicuddy, and Ross Sea circulation simulations from Stefanie Mack are graciously appreciated. We thank Marino Vacchi for providing unpublished CCAMLR documents CONCLUSIONS describing some of the Italian Antarctic cruises. A special thanks to Ari Friedlaender for the guidance and enthusiasm throughout this research. L.B.D. thanks Julie Morgan for the The continental shelf break/slope region of the introduction to CCPO. This research was the basis for Ross Sea provides a productive habitat with features L.B.D.’s MSc thesis. that favor Antarctic krill. Similarly, the inner shelf region within Terra Nova Bay provides a habitat that LITERATURE CITED supports crystal krill and the early life stages of Antarctic silverfish. The localization of energy trans- Ackley SF, Bengston JL, Boveng P, Castellini M and others fer in these regions provides a defined area for quan- (2003) A top-down, multidisciplinary study of the struc- ture and function of the pack-ice ecosystem in the east- tifying energy flows in the food web. This localization ern Ross Sea, Antarctica. Polar Rec 39: 219−230 also makes the food web productivity vulnerable to Ainley DG, Ballard G, Dugger KM (2006) Competition environmental changes. among penguins and cetaceans reveals trophic cascades This study takes a bottom-up approach to defining in the western Ross Sea, Antarctica. Ecology 87:2080−2093 habitat characteristics that are based on the assump- Ainley DG, Russell JL, Jenouvrier S, Woehler EJ, Lyver PO, Fraser WR, Kooyman GL (2010) Antarctic penguin tion that the distributions of the 3 mid-trophic level response to habitat change as Earth’s troposphere species are controlled by environmental conditions. reaches 2°C above preindustrial levels. Ecol Monogr 80: However, predation pressure and predator foraging 49−66 strategies are important top-down controlling factors Ainley DG, Ballard G, Jones RM, Jongsomjit D, Pierce SD, Smith WO Jr, Veloz S (2015) Trophic cascades in the that need to be combined with environmental controls western Ross Sea, Antarctica: revisited. Mar Ecol Prog to account for variability in prey distributions. Prey Ser 534: 1−16 behavior such as vertical migrations and swarm for- Amakasu K, Ono A, Hirano D, Moteki M, Ishimaru T (2011) mations are also biological and behavioral compo- Distribution and density of Antarctic krill (Euphausia superba) and ice krill (E. crystallorophias) off Adélie nents that may play a role in structuring the distribu- Land in austral summer 2008 estimated by acoustical tions of the 3 mid-trophic level species (Taki et al. methods. Polar Sci 5: 187−194 2008, Wiebe et al. 2011). The distribution of early Arrigo KR, Weiss AM, Smith WO Jr (1998a) Physical forcing life stages is likely controlled by bottom up factors of phytoplankton dynamics in the southwestern Ross such as circulation and temperature (e.g. Hofmann & Sea. J Geophys Res 103:1007−1021 Arrigo KR, Worthen D, Schnell A, Lizotte MP (1998b) Pri- Hüsrevo˘ glu 2003, Piñones & Federov 2016, Murphy et mary production in Southern Ocean waters. J Geophys al. 2017), but older life stages are more likely con- Res 103: 15587−15600 trolled by growth potential and predation (Atkinson et Arrigo KR, Van Dijken GL, Bushinsky S (2008) Primary pro- al. 2008). Refinement of the habitats identified in this duction in the Southern Ocean, 1997–2006. J Geophys Res 113: C08004 study requires inclusion of the effects of predation, as Ashjian CJ, Rosenwaks GA, Wiebe PH, Davis CS and others well as behavioral and species interactions, such as (2004) Distribution of zooplankton on the continental competition, both within and between trophic levels. shelf off Marguerite Bay, Antarctic Peninsula, during aus- The habitats for the 3 mid-trophic level species tral fall and winter, 2001. Deep Sea Res II 51:2073−2098 Atkinson A, Whitehouse MJ, Priddle J, Cripps GC, Ward P, identify areas of the Ross Sea that deserve focused Brandon MA (2001) South Georgia, Antarctica: a produc- management. Environmental changes in these re- tive, cold water, pelagic ecosystem. Mar Ecol Prog Ser gions coupled with possible pressures from resource 216: 279−308 use, such as fishing, could alter the balance in the food Atkinson A, Siegel V, Pakhomov EA, Rothery P (2004) Long- web and overall production of the Ross Sea. The re- term decline in krill stock and increase in salps within the Southern Ocean. Nature 432: 100−103 sults from this study provide further rationale and Atkinson A, Siegel V, Pakhomov EA, Rothery P and others support for ecosystem-level conservation that underlie (2008) Oceanic circumpolar habitat of Antarctic krill. the selection of the Ross Sea marine protected areas. Mar Ecol Prog Ser 362:1−23 Davis et al.: Krill, silverfish, and environmental correlations 63

Azzali M, Kalinowski J (2000) Spatial and temporal distribu- Fevolden SE (1980) Krill off Bouvetöya and in the southern tion of krill Euphausia superba biomass in the Ross Sea Weddell Sea with a description of larval stages of (1989-1990 and 1994). In: Faranda FM, Guglielmo L, Euphausia crystallorophias. Sarsia 65: 149−162 Ianora A (eds) Ross Sea ecology. Springer-Verlag, Berlin, Getis A, Ord JK (1992) The analysis of spatial association by p 433−455 use of distance statistics. Geogr Anal 24: 189−206 Azzali M, Leonori I, De Felice A, Russo A (2006) Spatial-tem- Granata A, Guglielmo L, Greco S, Vacchi M, Sidoti O, poral relationships between two euphausiid species in Zagami G, La Mesa M (2000) Spatial distribution and the Ross Sea. Chem Ecol 22:S219−S233 feeding habits of larval and juvenile Pleuragramma Ballard G, Jongsomjit D, Veloz SD, Ainley DG (2012) Coex- antarcticum in the western Ross Sea (Antarctica). In: istence of the mesopredators in an intact polar ocean Faranda FM, Guglielmo L, Ianora A (eds) Ross Sea ecol- ecosystem: the basis for defining a Ross Sea marine pro- ogy. Springer-Verlag, Berlin, p 369−393 tected area. Biol Conserv 156: 72−82 Guglielmo L, Granata A, Greco S (1997) Distribution and Brinton E, Townsend AW (1991) Development rates and abundance of postlarval and juvenile Pleuragramma habitat shifts in the Antarctic neritic euphausiid Euphausia antarcticum (Pisces, Nototheniidae) off Terra Nova Bay crystallorophias, 1986−87. Deep-Sea Res 38: 1195−1211 (Ross Sea, Antarctica). Polar Biol 19: 37−51 Cavalieri D, Parkinson C, Gloersen P, Zwally HJ (1996) Sea Guglielmo L, Donato P, Zagami G, Granata A (2009) Spatio- ice concentrations from Nimbus-7 SMMR and DMSP temporal distribution and abundance of Euphausia crys- SSM/I passive microwave data. National Snow and Ice tallorophias in Terra Nova Bay (Ross Sea, Antarctica) Data Center, Boulder, CO. http://nsidc.org/data/nsidc- during austral summer. Polar Biol 32: 347−367 0051.html Haberman KL, Ross RM, Quetin LB (2003) Diet of the Daly KL (1990) Overwintering development, growth, and Antarctic krill (Euphausia superba Dana): II. Selective feeding of larval Euphausia superba in the Antarctic grazing in mixed phytoplankton assemblages. J Exp Mar marginal ice zone. Limnol Oceanogr 35: 1564−1576 Biol Ecol 283:97−113 Daly KL (2004) Overwintering growth and development of Hofmann EE, Hüsrevo˘glu YS (2003) A circumpolar model- larval Euphausia superba: an interannual comparison ing study of habitat control of Antarctic krill (Euphausia under varying environmental conditions west of the superba) reproductive success. Deep-Sea Res II 50: Antarctic Peninsula. Deep Sea Res II 51:2139−2168 3121−3142 Daly KL, Macaulay MC (1988) Abundance and distribution Hofmann EE, Capella JE, Ross RM, Quetin LB (1992) Models of krill in the ice edge zone of the Weddell Sea, austral of early life history of Euphausia superba −Part 1. Time spring 1983. Deep-Sea Res 35: 21−41 and temperature dependence during the descent-ascent Daly KL, Macaulay MC (1991) Influence of physical and bio- cycle. Deep-Sea Res 39: 1177−1200 logical mesoscale dynamics on the seasonal distribution Hosie GW (1991) Distribution and abundance of euphausiid and behavior of Euphausia superba in the antarctic mar- larvae in the Prydz Bay region, Antarctica. Antarct Sci 3: ginal ice zone. Mar Ecol Prog Ser 79: 37−66 167−180 Daly KL, Zimmerman JJ (2004) Comparisons of morphol- Hosie GW (1994) The macrozooplankton communities in the ogy and neritic distributions of Euphausia crystallo- Prydz Bay region, Antarctica. In: El-Sayed SZ (ed) South- rophias and Euphausia superba furcilia during autumn ern Ocean ecology: the BIOMASS perspective. Cam- and winter west of the Antarctic Peninsula. Polar Biol bridge University Press, Cambridge, p 93–123 28:72−81 Hubold G (1984) Spatial distribution of Pleuragramma ant - Daneri GA, Carlini AR (2002) Fish prey of southern elephant arcticum (Pisces: Nototheniidae) near the Filchner- and seals, Mirounga leonine, at King George Island. Polar Larsen Ice Shelves (Weddell Sea/Antarctica). Polar Biol Biol 25: 739−743 3: 231−236 DeWitt HH (1970) The character of the midwater fish fauna Ichii T (1990) Distribution of Antarctic krill concentrations of the Ross Sea, Antarctica. In: Holdgate MW (ed) exploited by Japanese krill trawlers and minke whales. Antarctic ecology, Vol 1. Academic Press, New York, NY, Proc NIPR Symp Polar Biol 3:36−56 p 299–340 Kaufman DE, Friedrichs MAM, Smith WO Jr, Hofmann EE, DeWitt HH, Heemstra PC, Gon O (1990) Nototheniidae. In: Dinniman MS, Hemmings JC (2017) Climate change Gon O, Heemstra PC (eds) Fishes of the Southern Ocean. impacts on southern Ross Sea phytoplankton composi- JLB Smith Institute of Ichthyology, Grahamstown, tion, productivity, and export. J Geophys Res Oceans p 279–331 122: 2339−2359 Dinniman MS, Klinck JM, Smith WO (2003) Cross-shelf Kellermann A (1986a) Geographical distribution and abun- exchange in a model of the Ross Sea circulation and bio- dance of post larval and juvenile Pleuragramma ant - geochemistry. Deep Sea Res II 50: 3103−3120 arcticum (Pisces, Notothenioidei) off the Antarctic Penin- Donnelly J, Torres JJ (2008) Pelagic fishes in Marguerite sula. Polar Biol 6: 111−119 Bay region of the West Antarctic Peninsula shelf. Deep Kellermann A (1986b) On the biology of early life stages of Sea Res II 55:523−539 notothenioid fishes (Pisces) off the Antarctic Peninsula. Donnelly J, Torres JJ, Sutton TT, Simoniello C (2004) Fishes Ber Polarforsch 31: 1−155 of the eastern Ross Sea, Antarctica. Polar Biol 27:637−650 Koubbi P, O’Brien C, Loots C, Giraldo C and others (2011) Elith J, Leathwick JR (2009) Species distribution models: Spatial distribution and inter-annual variations in the ecological explanation and prediction across space and size frequency distribution and abundances of Pleura- time. Annu Rev Ecol Evol Syst 40:677−697 gramma antarcticum larvae in the Dumont d’Urville Sea Emslie SD, McDaniel JD (2002) Adélie penguin diet and cli- from 2004 to 2010. Polar Sci 5: 225−238 mate change during the middle to late Holocene in La Mesa M, Eastman JT (2012) Antarctic silverfish: life northern Marguerite Bay, Antarctic Peninsula. Polar Biol strategies of a key species in the high-Antarctic ecosys- 25: 222−229 tem. Fish Fish 13: 241−266 64 Mar Ecol Prog Ser 584: 45–65, 2017

La Mesa M, Catalano B, Russo A, Greco S, Vacchi M, Azzali macrura in Gerlache Strait and Crystal Sound, Antarc- M (2010) Influence of environmental conditions of spatial tica. Mar Ecol Prog Ser 109: 131−142 distribution and abundance of early life stages of Antarc- O’Brien DP (1987) Direct observations of the behavior of E. tic silverfish, Pleuragramma antarcticum (Nototheni- superba and E. crystallorophias (Crustacea: Euphausi- idae), in the Ross Sea. Antarct Sci 22: 243−254 acea) under pack ice during the Antarctic spring of 1985. Laws RM (1985) The ecology of the Southern Ocean: the J Crustac Biol 7: 437−448 Antarctic ecosystem, based on krill, appears to be mov- O’Driscoll RL, Macaulay GJ, Gauthier S, Pinkerton M, ing toward a new balance of species in its recovery from Hanchet S (2011) Distribution, abundance and acoustic the inroads of whaling. Am Sci 73: 26−40 properties of Antarctic silverfish (Pleuragramma antarc - Lawson GL, Wiebe PH, Ashjian CJ, Gallager SM, Davis CS, ticum) in the Ross Sea. Deep-Sea Res II 58: 181−195 Warren JD (2004) Acoustically-inferred zooplankton dis- Orsi AH, Wiederwohl CL (2009) A recount of Ross Sea tribution in relation to hydrography west of the Antarctic waters. Deep-Sea Res II 56:778−795 Peninsula. Deep-Sea Res II 51: 2041−2072 Pakhomov EA, Perissinotto R (1996) Antarctic neritic krill Lawson GL, Wiebe PH, Ashjian CJ, Stanton K (2008) Euphausia crystallorophias: spatio-temporal distribution, Euphausiid distribution along the Western Antarctic Pe - growth and grazing rates. Deep-Sea Res I 43: 59−87 ninsula — Part B: distribution of euphausiid aggregations Pakhomov EA, Perissinotto R (1997) Spawning success and and biomass, and associations with environmental fea- grazing impact of Euphausia crystallorophias in the tures. Deep-Sea Res II 55: 432−454 Antarctic shelf region. In: Battaglia B, Valencia J, Walton Leonori I, De Felice A, Canduci G, Costantini I, Biagiotti I, DWH (eds) Antarctic communities: species, structure Giuliani G, Budillon G (2017) Krill distribution in relation and survival. Cambridge University Press, Cambridge, to environmental parameters in mesoscale structures in p 187–192 the Ross Sea. J Mar Syst 166:159−171 Pinkerton MH, Bradford-Grieve JM (2014) Characterizing Mack SL, Dinniman MS, McGillicuddy DJ Jr, Sedwick PN, foodweb structure to identify potential ecosystem effects Klinck JM (2017) Dissolved iron transport pathways in of fishing in the Ross Sea, Antarctica. ICES J Mar Sci 71: the Ross Sea: influence of tides and mesoscale eddies in 1542−1553 a regional ocean model. J Mar Syst 166:73−86 Pinkerton MH, Bradford-Grieve JM, Hanchet SM (2010) A Mackintosh NA (1973) Distribution of postlarval krill in the balanced model of the food web of the Ross Sea, Antarc- Antarctic. Discov Rep 36: 95−156 tica. CCAMLR Sci 17: 1−31 Marr JWS (1962) The natural history and geography of the Piñones A, Fedorov AV (2016) Projected changes of Antarc- Antarctic krill (Euphausia superba Dana). Discov Rep 32: tic krill habitat by the end of the 21st century. Geophys 33−464 Res Lett 43: 8580−8589 McGillicuddy DJ Jr, Budillon G, Kustka A (2017) Mesoscale Piñones A, Hofmann EE, Dinniman MS, Davis LB (2016) and high-frequency variability in the Ross Sea (Antarc- Modeling the transport and fate of euphausiids in the tica): an introduction to the special issue. J Mar Syst 166: Ross Sea. Polar Biol 39:177−187 1−3 Preisendorfer RM, Mobley CD (1988) Principal component Miller RG (1993) A history and atlas of the fishes of the analysis in meteorology and oceanography. Develop- Antarctic Ocean. Foresta Institute for Ocean and Moun- ments in Atmospheric Science, Vol 17. Elsevier, New tain Studies, Carson City, NV York, NY Miquel JC (1991) Distribution and abundance of post-larval Priddle J, Croxall JP, Everson I, Heywood RB, Murphy EJ, krill (Euphausia superba Dana) near Prydz Bay in sum- Prince PA, Sear CB (1988) Large-scale fluctuations in dis- mer with reference to environmental conditions. Antarct tribution and abundance of krill — a discussion of possi- Sci 3: 279−292 ble causes. In: Sahrhage D (ed) Antarctic Ocean and Murase H, Kitakado T, Hakamada T, Matsuoka K, Nishi- resources variability. Springer, Berlin, p 169–182 waki S, Naganobu M (2013) Spatial distribution of Quetin LB, Ross RM (1984) School composition of the Antarctic minke whales (Balaenoptera bonaerensis) in Antarctic krill Euphausia superba in the waters west of relation to spatial distributions of krill in the Ross Sea, the Antarctic Peninsula in the austral summer of 1982. Antarctica. Fish Oceanogr 22: 154−173 J Crustac Biol 4: 96−106 Murphy EJ, Watkins JL, Trathan PN, Reid K and others Ross RM, Quetin LB, Kirsch E (1988) Effect of temperature (2007) Spatial and temporal operation of the Scotia Sea on development times and survival of early larval stages ecosystem: a review of large-scale links in a krill- of Euphausia superba Dana. J Exp Mar Biol Ecol 121: centered food web. Philos Trans R Soc B 362:113−148 55−71 Murphy EJ, Thorpe SE, Tarling GA, Watkins JL, Fielding S, Sala A, Azzali M, Russo A (2002) Krill of the Ross Sea: distri- Underwood P (2017) Restricted regions of enhanced bution, abundance and demography of Euphausia growth of Antarctic krill in the circumpolar Southern superba and Euphausia crystallorophias during the Ocean. Sci Rep 7: 6963 Italian Antarctic Expedition (January–February 2000). Naganobu M, Nishiwaki S, Yasuma H, Matsukura R and Sci Mar 66:123−133 others (2006) Interactions between oceanography, krill Siegel V (2005) Distribution and population dynamics of and baleen whales in the Ross Sea and adjacent waters: Euphausia superba: summary of recent findings. Polar an overview of Kaiyo Maru-JARPA joint survey on Biol 29: 1−22 2004/05. J Cetacean Res Manag 10: 438−439 Siegel V, Loeb V (1995) Recruitment of Antarctic krill Nicol S (2006) Krill, currents, and sea ice: Euphausia (Euphausia superba) and possible causes for its variabil- superba and its changing environment. Bioscience 56: ity. Mar Ecol Prog Ser 123: 45−56 111−120 Smith WO Jr, Ainley DG, Cattaneo-Vietti R (2007) Trophic Nordhausen W (1994) Winter abundance and distribution of interactions within the Ross Sea continental shelf ecosys- Euphausia superba, E. crystallorophias and Thysanoessa tem. Philos Trans R Soc Lond B Biol Sci 362: 95−111 Davis et al.: Krill, silverfish, and environmental correlations 65

Smith WO Jr, Sedwick PN, Arrigo KR, Ainley DG, Orsi AH gramma antarcticum in Terra Nova Bay, Ross Sea. (2012) The Ross Sea in a sea of change. Oceanography Antarct Sci 16: 299−305 25: 90−103 Vacchi M, DeVries AL, Evans CW, Bottaro M, Ghigliotti L, Smith WO Jr, Ainley DG, Arrigo KR, Dinniman MS (2014a) Cutroneo L, Pisano E (2012a) A nursery area for the The oceanography and ecology of the Ross Sea. Annu Antarctic silverfish Pleuragramma antarcticum at Terra Rev Mar Sci 6: 469−487 Nova Bay (Ross Sea): first estimate of distribution and Smith WO Jr, Dinniman MS, Hofmann EE, Klinck JM abundance of eggs and larvae under the seasonal sea- (2014b) The effects of changing winds and temperatures ice. Polar Biol 35: 1573−1585 on the oceanography of the Ross Sea in the 21st century. Vacchi M, Koubbi P, Ghigliotti L, Pisano E (2012b) Sea-ice Geophys Res Lett 41:1624−1631 interactions with polar fish: focus on the Antarctic silver- Spiridonov VA (1996) A scenario of the Late-Pleistocene- fish life history. In: Prisco G, Verde C (eds) Adaption and Holocene changes in the distributional range of Antarc- evolution in marine environments, Vol 1: from Pole to tic krill (Euphausia superba Dana). Polar Biol 15: 161−174 Pole. Springer-Verlag, Berlin, p 51–73 Stirling I (1969) Distribution and abundance of the Weddell Weber LH, El-Sayed SZ (1985) Spatial variability of phyto- seal in the western Ross Sea, Antarctica. NZ J Mar plankton and the distribution and abundance of krill in Freshw Res 3: 191−200 the Indian sector of the Southern Ocean. In: Siegfried Sushin VA, Shulgovsky KE (1999) Krill distribution in the WR, Condy PR, Laws RM (eds) Antarctic nutrient cycles western Atlantic sector of the Southern Ocean during and food webs. Springer, Berlin, p 284–293 1983/84, 1984/85 and 1987/88 based on the results of Weber LH, El-Sayed SZ, Hampton I (1986) The variance soviet meso-scale surveys conducted using an Isaacs-Kid spectra of phytoplankton, krill and water temperature in Midwater trawl. CCAMLR Sci 6:59−70 the Antarctic Ocean south of Africa. Deep-Sea Res A 33: Taki K, Yabuki T, Noiri Y, Hayashi T, Naganobu M (2008) 1327−1343 Horizontal and vertical distribution and demography of White MG, Piatkowski U (1993) Abundance, horizontal and euphausiids in the Ross Sea and its adjacent waters in vertical distribution of fish in the Weddell Sea micronek- 2004/2005. Polar Biol 31: 1343−1356 ton. Polar Biol 13: 41−53 Thomas PG, Green K (1988) Distribution of Euphausia crys- Wiebe PH, Ashjian CJ, Lawson GL, Piñones A, Copley NJ tallorophias within Prydz Bay and its importance to the (2011) Horizontal and vertical distribution of euphausiid inshore marine ecosystem. Polar Biol 8: 327−331 species on the western Antarctic Peninsula U.S. GLOBEC Trathan PN, Brierley AS, Brandon MA, Bone D and others Southern Ocean study site. Deep-Sea Res II 58:1630−1651 (2003) Oceanographic variability and changes in Ant- Witek Z, Kalinowski J, Grelowski A, Wolnomiejski N (1981) arctic krill (Euphausia superba) abundance at South Studies of aggregations of krill (Euphausia superba). Georgia. Fish Oceanogr 12:569−583 Meeresforsch 28:228−243 Trunov IA (2001) Occurrence of Pleuragramma antarcticum Witek Z, Kalinowski J, Grelowski A (1988) Formation of (Nototheniidae) off South Georgia Island and the South Antarctic krill concentrations in relation to hydrody- Sandwich Islands (Antarctica). J Ichthyol 41: 549−550 namic processes and social behavior. In: Sahrhage D (ed) Vacchi M, La Mesa M, Greco S (1999) Summer distribution Antarctic Ocean and resource variability. Springer, and abundance of larval and juvenile fishes in the west- Berlin, p 237–244 ern Ross Sea. Antarct Sci 11: 54−60 Woehler EJ (1993) The distribution and abundance of Ant- Vacchi M, La Mesa M, Dalù M, Macdonald J (2004) Early arctic and Subantarctic penguins. Scientific Committee life stages in the life cycle of Antarctic silverfish, Pleura- for Antarctic research, Cambridge

Editorial responsibility: Marsh Youngbluth, Submitted: May 12, 2017; Accepted: September 25, 2017 Fort Pierce, Florida, USA Proofs received from author(s): November 20, 2017